Every so often, the sun has an epic tantrum — and if Earth’s in the way, bad things can happen. That’s why NASA and Amazon are teaming up with a new machine learning application to better predict the coming of a solar “superstorm.”
We’ve seen these superstorms cause damage in the past. In 1859, an incident called the Carrington event reportedly disrupted telegraph communications. And in winter 1989, thousands of residents of Quebec, Canada were plunged into darkness when their hydro cut out. While these worst events are rare, on average they are expected every 50 years.
These incidents are due to eruptions on the sun. The sun periodically sends out flares (which you can see in large telescopes) and often, these flares are accompanied by bursts of invisible radiative particles called coronal mass ejections. Usually Earth’s magnetic field protects against the radiation. But occasionally, there are so many particles that they can produce huge solar storms, knocking out satellites, power lines and other vital infrastructure for humans.
So NASA and Amazon together are working on machine learning applications. It’s not an easy task, Amazon said in a blog post.
“Given just how rare superstorms are, there are very few historical examples that can be used to train algorithms. This makes common machine learning approaches like supervised learning woefully inadequate for predicting superstorms,” Amazon stated. “Additionally, with dozens of past and current satellites gathering space weather information from different key vantage points around Earth, the amount of data is prodigious — and the attempt to find correlations laborious when searched conventionally.”
Amazon’s AWS Professional Services and Machine Learning Solutions Lab have another approach. They’re using both unsupervised learning and anomaly prediction to better predict the conditions that are associated with superstorms. AWS is able to examine as many as 1,000 data sets simultaneously, based on rankings of anomalies (generated by NASA) to find patterns that are unique to superstorms.
Before long, NASA and Amazon plan to offer a data “lake” (or repository with reams of raw data) to let researchers crunch the numbers for themselves. The plan is to make forecasting even stronger by looking at the anomalies and making simulations about current-day superstorms — and the extreme ones of history, like the Carrington event.
“There’s a lot of data, and factors like time lags add to the complexity,” said lead researcher Janet Kozyra, a heliophysicist (sun researcher) at NASA, in the statement. “With Amazon, we can take every single piece of data that we have on superstorms, and use anomalies we have detected to improve the models that predict and classify superstorms effectively.”